The two-way reading hypothesis

Listen to audio

Beyond the visual word recognition stage, how does reading continue? Most current models consider that information processing takes two parallel paths. One leads to access to the lexicon and meaning of words, the other to their phonological representation.

Neuropsychology (Marshall & Newcombe, 1973) provided the first clues to this two-way organization: in addition to peripheral alexias, which include pure alexia, neuropsychology distinguishes at least one form of deep or central alexia, in which the patient has partial access to the meaning of the word without recovering its pronunciation, and a form of surface alexia, in which errors reflect over-regularization with the application of grapheme-phoneme conversion rules without understanding the word.

Marshall and Newcombe, then Coltheart, proposed that these deficits reflect the selective deterioration of at least two reading pathways, a surface pathway that enables grapheme-phoneme conversion of all character strings, and a deep pathway that retrieves the lexical-semantic information associated with each word. According to this model, various factors modulate the use of one or the other pathway. Three essential categories of strings can be distinguished :

pseudowords such as "bakifo" (surface path only) ;
regular words such as "bateau" (surface and deep path) ;
irregular words such as "onion" (deep path only).

These theoretical distinctions introduced by neuropsychology have subsequently had an important influence on research in the normal subject. In psycholinguistics, the effects of pseudo-homophony, homophony, phonological priming and phonological consistency have suggested that the two reading pathways function in parallel, and that phonological information is systematically activated at a sub-liminal level during reading. These findings have been incorporated into detailed models, an excellent critical synthesis of which is presented by Perry et al. (2007).

Neuroimaging research then focused on dissecting the brain areas associated with the two reading "routes". Two types of method have been used to modulate passage through the lexical or grapheme-phoneme conversion pathways. In the "variable-stimulus" method, stimuli are contrasted which, according to theory, take different routes (e.g. word reading versus pseudoword reading; regular versus irregular words; frequent (stored) versus rare words (little or not encoded in the lexicon)). The "constant-stimulus, variable-task" method contrasts tasks that focus the subject's attention on one level of phonological processing (e.g. rhyme judgment) or lexical processing (synonymy judgment).

The results of these different methods are beginning to converge, although many unexplained differences remain. A recent review (Jobard, Crivello, & Tzourio-Mazoyer, 2003) implicates the back of the middle temporal gyrus, the anterior ventral aspect of the temporal lobe and the triangular part of the inferior frontal region in the deep lexical-semantic pathway, and the left superior and middle temporal regions as well as the supramarginal gyrus and the opercular part of the inferior frontal region in the surface grapheme-phoneme conversion pathway. These findings, most often based on fMRI or PET scans, are confirmed by a high temporal resolution MEG study (Simos et al., 2002). This suggests that the two pathways function in parallel, but that reading speed is determined by one or the other pathway, depending on the nature of the word being read.